Kakuma Refugee Camp: Household Vulnerability Study

Transcription

1 Kakuma Refugee Camp: Household Vulnerability Study Dr. Helen Guyatt Flavia Della Rosa Jenny Spencer Dr. Eric Nussbaumer Perry Muthoka Mehari Belachew

2 Acknowledgements Commissioned by WFP, UNHCR and partners in Kakuma Camp Technical Steering Committee, particularly Yvonne Forsen Key informants who offered their time and insights Survey team: coordinators, enumerators and interpreters Refugee households and community leaders

3 Aims After 20 years not all refugees have the same humanitarian assistance needs Livelihood opportunities, income sources, differences in socio-economic vulnerability Feasibility of targeting assistance based on actual needs and vulnerability Gold standard calculations for each HH of cash equivalent consumption expenditure per capita per day Evaluation of different targeting approaches: inclusion and exclusion errors

4 Methods

5 Scoping exercise Qualitative and contextual information from literature review, stakeholder interviews, focus-group discussions and HH visits Secondary quantitative data from UNHCR, WFP and other partners HH questionnaire and sampling methodology

6 Sampling Administrative units: Sub-camp, Zones, Blocks. 126 blocks across 12 zones 500 HHs/sub-camp equally divided across the blocks in that camp Camp Blocks HHs/block K K K K HHs (13,378 people)

7 Every single block in Kakuma (126) was sampled. First time such a comprehensive HH survey has been done in Kakuma.

8 HH survey Training workshop (2-6 November) Fieldwork (7 November 4 December) Questionnaire sections: 1. HH details 2. HH roster 3. Housing and wealth indicators 4. Livelihoods and income 5. Food assistance, Bamba Chakula, NFIs 6. HH consumption expenditure 7. HH coping strategies

9 HH survey results Demographics

10 Distribution of the main CoOs Similar distribution to UNHCR Captured all ethnic groups South K1: 53% Sudanese SS, 27% somali 52% K2: 50% somali, 17% SS South Somali K2 K3: Sudanese 55% somali, 24% 50% SS 17% K4: 91% SS, 0% somali South K3 Sudanese PUT this as a chart? 24% Somali 56% Somali 27% Other 21% Other 33% Other 20% K4 South Sudanese 91% Somali 0% Other 9%

11 Year of arrival K1, 10% K2, 7% K3, 13% All new arrivals since 2014 K4, 70%

12 Definition of a HH UNHCR: Ration card Survey: Persons who both live and eat together Range from 1-30 members Median HH size: K1: 7.3 K2: 6.8 K3: 6.9 K4: 5.8

13 HH size 1 Total UNHCR Database 33% Survey (if assume ration card is a HH: 2,838 HHs) Survey (our HH definition: 2,000 HHs) 22% 5% Of the 22% RC size 1, 82% have joined other HHs Of the 1,898 HHs >size 1 28% have more than one RC 17% have had a RC size 1 join them Of the true HH size 1: about 2/3 are young men under 30 though rates of employment and business ownership are lower than for the total population.

14 Examples of joining up males aged between 20 and 28 in K3. Sudanese and recent arrivals (2013). Live and eat together with 9 ration cards between them. 2. HH with 6 members all brothers and sisters with 3 ration cards. In K1 from Uganda. Oldest brother arrived in 2013 on one ration card and then younger siblings followed on other ration cards.

15 HH survey results Wealth assets

16 Wealth assets 77% of HHs own a mobile phone >80% in K1, K2, and K3 56% in K4 Key wealth assets: bicycle, TV, wheelbarrow, dining table, solar panels 18% have 1 of these (8% in K4) 12% of HHs have at least 2/5 (2% in K4)

17 Electricity 21% have electricity (85% from community generator) K1: 31% K4: <1% Somalis and Ethiopians: Nearly half South Sudanese: 2%

18 HH survey results Livelihoods

19 Legal constraints to livelihoods Limited freedom of movement No access to land for agricultural production No access to the credit or saving sector Restrictions on livestock ownership Can apply for a business license but not a work permit Not allowed to travel for business purposes

20 Previous livelihoods Main HH activities before arrival: farming (43%) and livestock rearing (5%) Unskilled labor (22%) Business (8%) Feasible now? Block leaders do not consider farming a sustainable activity due to the harsh climatic conditions in Turkana.

21 Vocational training 6% of adults (age 15-64) have received vocational training 10% of English speakers vs. 4% of non-english speakers

22 Current livelihoods Only 20% of HHs have at least one member with some type of employment (includes business; only 6% in K4) Incentive (43% of those with employment; 9% of all HHs) Business (40% of those with employment; 8% of all HHs) Mostly K1-K3, only 2% in K4 Most arrived before 2011 Predominantly shops/kiosks/hawkers Over half operating for one year or less

23 Proportion of HHs sampled reporting cash income in last 30 days Employment 10.4% Reselling ration 9.6% Running a business Small jobs (petty trading) 8.2% 8.0% Gifts from friends/relatives outside camp Gifts from friends/relatives inside camp Selling other items 2.0% 1.9% 6.0% 68% had no income; 25% from one source only. Only 9% of HHs with an income (3% of all HHs) received >10,000 Ksh

24 Cash incomes from each source in last 30 days

25 HH survey results Measuring vulnerability

26 Vulnerability = food insecurity and poverty **A highly vulnerable HH would not be able to provide for itself in the absence of, or with cuts to, assistance

27 Indicator = Cash equivalent consumption expenditure per capita per day (CECE pc pd) **Excludes all gifted and assistance Includes purchased, own production and in stock food Includes both consumable and durable NFIs

28 CECE pc pd (n=1986) 7% HHs had none (45% for food only) Food, 63% For all HHs, median was 6.39 Ksh (mean of 17.90) Variations by subcamp, HH size, female-headed etc. Durable NFIs, 7% Consumable NFIs, 30%

29 CECE pc pd 6.39 median (17.9 mean); range from 0 to Ksh Frequency Ksh More Max 1,260

30 CECE pc pd compared to a set of vulnerability thresholds 31% can provide own NFIs (15 Ksh) More Frequency 15% can provide ½ own food (31 Ksh) 9.1% can provide own NFIs & ½ own food (46 Ksh) 5.7% can provide own food (62 Ksh) 4.2% can provide own food & NFIs (77 Ksh) 1.7%: above Kenyan poverty line (125 Ksh) Ksh.

31 Proportion NOT vulnerable Elimination of all assistance (77 Ksh per capita) Reduction in food assistance by half (31 Ksh per capita) Elimination of NFI assistance (15 Ksh per capita) Total 4.2% 15% 31% K1 7.6% 26% 41% K4 1.8% 4.6% 14% Ethiopian 15% 35% 57% Somali 7.3% 27% 50% South Sudanese 1.2% 4.4% 12%

32 Who are the non-vulnerables? Vulnerables? 0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100% K1 K2 K3 K4 Somalia South Sudan Ethiopia Arrived before 2014 Arrived in 2014 or later *Based on total basket of 77 Ksh

33 Who are the non-vulnerables? Vulnerables? 0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100% Male headed Female headed HH size 1 HH size 2-5 HH size 6-10 HH size >10 No business Has business *Based on total basket of 77 Ksh

34 Vulnerability Very few could survive without assistance

35 Targeting

36 Targeting Effectiveness: Inclusion/exclusion errors Costs: Implementation vs. savings Categorical Proxy means testing Community-based targeting Self-targeting

37 Headline goes here Effectiveness: Inclusion and exclusion errors Inclusion error (leakage rate) = proportion of those selected for assistance who don t need it = false positives. all positives (false positives and true positives) Exclusion error (undercoverage rate) = proportion of those who need assistance but which are excluded = false negatives. all those in need (false negatives and true positives)

38 Costs: Screening costs for targeting approaches Means testing ~US$2.74m ($100 per 27,352 HHs) Proxy means testing (PMT) ~1/2 Categorical targeting (CT) ~5% Community based targeting (CBT) ~1%

39 Targeting Categorical

40 Categorical targeting: Effectiveness Elimination of all assistance (77 Ksh) HHs targeted False positives Inclusion error False negatives Exclusion error Status quo % 0 0% Female-headed % % Female, child, disabled or elderly headed % % HH with NO business % % HHs with NO business or incentive work % % Female-headed with NO business % % Reduction in food assistance by half (31 Ksh) HH with NO business % %

41 Costs and effectiveness of upscaling targeting to HHs with no business Value Calculations HHs targeted 25,017 27,352 (HHs camp) x (Prop. no business) HHs vulnerable 26,209 27,352 (HHs camp) x (Prop. vulnerable) HHs included by mistake 716 HHs targeted x inclusion error (0.0285) HHs excluded by mistake 1819 HHs vulnerable x exclusion error (0.0694) HHs not receiving assistance 2245 HHs camp-hhs targeted

42 Categorical targeting Inclusion/exclusion errors for business are promising But ~2,000 HHs (~12,000 people) will be wrongly excluded Does not fulfill do no harm

43 Targeting Proxy means testing

44 Proxy means testing 4 models 2 Regression: OLS and Elastic Net 2 Classifier: Logistic Regression and Extremely Random Trees (ERT) 2 datasets Comprehensive (23 variables, incl. wealth assets and income) Limited (12 variables, incl. location, CoO, year of arrival, HH size, head of HH, etc.) All vulnerability thresholds

45 Proxy means testing Limited dataset (HH=1980) Elimination of all assistance (77 Ksh) Reduction of food assistance by half (31 Ksh) OLS Regression Elastic Net Logistic Regression Extremely Random Trees Incl. Error 3.5% 3.9% 0.3% 0.4% Excl. Error 0.9% 0.5% 13% 2.6% Incl. Error 6.1% 6.4% 6.1% 6.9% Excl. Error 17% 15% 21% 12%

46 Proxy means testing Regression does not work for many model combinations ERT produces low errors but ~1,258 HHs (8,415 people) wrongly excluded

47 Targeting Community based targeting

48 Methods Further field work Interviewed community leaders of 123/126 blocks For the surveyed HHs: Knowledge of HHs Business and family ties Rankings: wealth assets, businesses, remittances and overall Ability to survive without assistance

49 Knowledge of HHs HHs known % blocks 50% 86% 75% 74% 90% 55% 100% 40% For ranking and analysis, only considered HHs they knew (1,602 HHs)

50 HH remittances and employment (incl. business) All HHs had no employment 16% Employment Uncertain of some or all employment 7% Knew all employment and could rank 76% All HHs had no remittances 32% Remittances Uncertain of some or all remittances 45% Knew all remittances and could rank 24%

51 Do block responses match HH survey responses? Employment (incl. business) Remittances Identified by at least one source 31% 16% Which source? CBT, 33% Survey, 40% Both, 27% CBT, 72% Survey, 20% Both, 7.5%

52 Community leaders ranking perception (Blocks=114) Comparison of overall wealth ranking by community leaders vs. actual ranking of consumption expenditure for the known HHs. Correlation determined using Spearman s rank correlation coefficient (1.0 indicates identical, -1.0 indicates exact opposite ranking) >1/3 had negative correlations. Only 3 blocks had strong correlations (>=0.8)

53 Which HHs could survive without assistance? 10% of available sample (1,599) identified by at least one source CBT, 61% Both, 4.3% Survey, 35% Only 37% of the 123 community leaders were able to identify at least one HH in their block that could survive without assistance

54 Community-based targeting FGDs Informal redistribution occurs: Wealthier HHs give to the poor Some communities share their food (eg. Dinkas from South Sudan) Community leaders are uncomfortable targeting out but would want to help identify vulnerables if they then got more assistance. If target in, vulnerables would include: widows, orphans, unaccompanied minors, foster children, single mothers, elderly people and sick people with chronic diseases

55 Community based targeting Many HHs are not known to the community leaders Rankings are poorly correlated with CECE pc pd

56 Targeting Self-targeting

57 Self-targeting out: Options discussed Businesses: Loans if directed at individuals; if allowed to leave Kakuma Incentive workers: Only if contracts and fixed prices Land in new camp: Climate too harsh for farming; fear of insecurity and conflict with local community

58 Self-targeting out: Concerns Link between identity and other services (i.e. health) with ration and ration card: If you are a refugee you must have your card. Fear that re-registration process will not work properly if business fails or HH loses employment. High risk aversion.

59 Self-targeting Most viable option Incentives may be outside donor control

60 Common truths, common myths Who is vulnerable?

61 Female-headed HHs: truly vulnerable 63% are South Sudanese 32% are in Kakuma 4 41% are recent arrivals At a disadvantage along almost every indicator (ability to earn, wealth assets, food security) CECE pc pd Not vulnerable, 77 Ksh Female 4.3 Female 2.0% Male 8.9 Male 6.3%

62 HH size 1: Not as vulnerable as previously thought 81% are male 42% in Kakuma 4 42% are new arrivals CECE pc pd Not vulnerable, 77 Ksh % HH size % 5.9% > >10 0.8%

63 New arrivals vs. established residents New arrivals: most vulnerable. Residents who have been in the camp longest are not necessarily least vulnerable (though low sample size n=48) CECE pc pd Before 1995

64 Findings and recommendations

65 Very few refugees can meet a significant proportion of their basic needs from their own resources. Without greater economic integration, the opportunities for targeting will remain limited.

66 1. Continue to provide full assistance to refugees 4.2% inclusion error and 0% exclusion error, do no harm policy Targeting options to eliminate assistance would either not work or exclude over 1,250 HHs in need. Self-targeting could be explored. 2. Conduct a HH census to update UNHCR statistics 3. Halt targeting of food assistance based on HH size -- HH size 1 are the least vulnerable HH size group

67 4. Conduct a needs assessment on vocational training requirements and to explore potential livelihoods -- English is an important factor in accessing vocational training -- Business and employment are not homogenous nor necessarily lucrative 5. NGO and donor organizations should work together to identify a common pay scale for incentive staff. 6. Remittances offer an important area for further research Potentially a very substantial income source but sensitive information

68 Thank You